BIOL 320 Lecture Notes - Lecture 11: Basal Ganglia, Biological Neural Network, Motor Learning
you need to explore motor space, generate variable actions
—
evaluate the outcome
—
[see slide]
—
E.G. walking, speaking, doing a cartwheel
—
reinforcement learning: this motor behaviours that you have that aren't innate but that aren't explicitly taught
•
it's sequential, you're learning multiple steps
—
once it becomes a habit, you don't need as much cortical input to keep on going
—
you need to basal ganglia to start doing that set of movements and then continue on doing it!
—
E.G. juggling
•
memorise target
1)
explore motor space
—
you initially make a lot of errors (ideally you start with bean bags, not fire); then you'll hit a set of
movements that you're brain will say
—
aha! correct!
2)
so correct movements are rewarded, and wrong movements are supressed
3)
there
are
genetic differences in propensity to learn, but it's not clear whether there's actual differences in the basal
ganglia that = more coordinated or less coordinated
•
there's a strong dopamine signal that comes from ventral tegmental and SN
•
conditioned stimulus
→
neurons can be trained to respond to the conditioned stimulus instead of the reward—so
you can train brain to get to the correct condition even before getting to your target
—
these neurons are nicely set up to complete these tasks
—
you can make link between the correct movement to the
desired end point
—
activity is bi
-
directional: you can also predict when unexpected things happen
•
so dopamine stamps in the correct movements
•
activity related to movement
→
motor-related response in the basal ganglia
—
neurons in basal ganglia are also good in directing bias in action (e.g.
Delong, 1970
):
•
when there's a reward present, there's a ranking depending on the motor behaviour
—
but that's shifted to some
degree when there's a reward
—
neurons are able to incorporate both pieces of information: motor movement & reward/bias
—
the biases in the basal ganglia are influenced by reward (
Hikosaka, 1998)
•
you need to show that ot's actively generated
—
you're purposely adding in variability instead of actually no being
good enough
—
you need to find specific basal ganglia loops or cells that are significant for behaviour, not neurons that are showing
—
so we've got components 2 & 3 pretty well-covered—but what about 1? motor exploration?
→
the challenge with
variability is that there's only a few studies that have looked at where it comes from
•
11. Basal Ganglia
February 14, 2018 11:38
BIOL 320 Page 1
you need to find specific basal ganglia loops or cells that are significant for behaviour, not neurons that are showing
variable activity because they're involved in smth completely different. and you're not going to have a specific
network dedicated
just
for juggling
—
you need to manipulate or record them
—
so studies in songbirds have been critical!
—
clear variability that has been determined to be actively generated
1)
there's a discrete neural circuit that can be observed and manipulated
2)
why?
•
there's all different kinds of BG networks we can study
—
for humans, ability to speak is super important!
•
there's only a handful of animals that have vocal learning: humans, cetaceans, hummingbirds, elephants, bats, songbirds!
•
sensory acquisition will shape the brain
—
sensorimotor learning
→
babbling an trial-and-error motor learning; start out pretty quiet and mumbly, but as he
practices it and compares it to the sound he's memorise and will either suppress or reward the activity! and so will
go back and forth in this process
—
at ~90 days old, will produce a crystallised song!
—
all songbirds learn their vocalisations very young: you have a tutor and a pupil, between 10
-
60 days old, pupil will listen to
song and will memorise in sensory acquisition phase
•
Songbirds
bird needs to make a template of the correct song, needs to remember it! and needs auditory feedback
—
needs to hear
the songs that they're making and be able to compare it to their memory
•
are vocal learners
—
have discrete circuit that can observed and manipulated
—
songbirds
•
it's like if we'd have a single loop for just hockey or cartwheels
—
in fact, we have an area devoted to speech!
—
so birds have BG and they use it to do all kinds of motor tasks (flying, eating, pecking, etc.) and then they also have a
specific cortical area devoted
just
for singing!
•
there's a motor pathway: HVC (= pre-motor area) & RA (= motor area)
→
this is necessary for song-production and
song
-
learning throughout life
—
anterior
-
forebrain pathway: necessary for song learning and plasticity, makes an indirect loop between HVC to RA
—
BG circuit:
•
so these are highly specialised! not involved in anything BUT singing
—
comparing songbirds to mammals
•
BIOL 320 Page 2
•
so there's similar loop receiving a whole bunch of dopamine
—
for the most part, avian and mammalian looks the same
—
but for some reason, area X has both striatum and
pallidal neurons in the same area
—
diagram of zebra finch song (very stereotyped song, always the same sequence):
•
but if you lesion the LMAN, birds will never learn a proper song
—
what's happening?? why?
•
lots of work is focused on adults because songs are more defined, so you can try to figure out what the BG is contributing
•
song (x = time, y = frequency)
—
so looking at adult plasticity
•
blue circle = baseline frequency. and every time the bird sings this specific syllable, we'll measure brain
activity
then, every time bird sings this syllable but sings a little bit higher than usual, he's hit with white noise (so it's
annoying, but not painful)
overtime, this causes bird to shift his song! shifting the song down so he can avoid the white noise
—
so adults
can adjust their song based on external variables!
BIOL 320 Page 3